This concept shares some similarities with <a target="_blank" rel="ugc" href="https://civarchive.com/models/1162484/blending-style">Blending</a> and <a target="_blank" rel="ugc" href="https://civarchive.com/models/2782139">Fixed-Point Camera</a>, but it focuses on using empty space as a deliberate and essential part of the composition—as if the image would not work without it.Version 0.0.0 was trained on a carefully selected dataset of 100 images in which empty space occupies more than half of the frame. As a result, applying this LoRA tends to increase the proportion of empty space within the composition. However, the effect becomes much more noticeable when using a trigger strength of 1.5 or higher, or when combined with prompts such as wide shot.Unlike my Fixed-Point Camera LoRA, this dataset was not strictly filtered to exclude subjects positioned near the center of the frame. If you prefer compositions where the subject is kept away from the center, I recommend using this LoRA together with my Fixed-Point Camera LoRA.

Negative space

Initial.dim4/alpha8, unet 1e-04, adamw8bit, warmup 100, 512² reso (262k px) 100images, cosine 600 steps

Files with SHA256: e4e2e34131cae0e6276215b0724de034c570a35d8b1dc07bdf41f6a8baf328c7